Estimating of Software Quality with Clustering Techniques

Software faults are one of major criteria to estimate the software quality or the software reliability. There is number of matrices defined that uses the software faults to estimate the software quality. When we have a large software system with thousands of class modules, then it is not easy to apply the software matrices on each module of software system. The present work is the solution of the defined problem. This paper aims at comparing different models based on clustering techniques: k-means (KM), fuzzy c-means (FCM) and hierarchical agglomerative clustering (HAC) for building software quality estimation system. We propose quality measure of partition clustering technique (KM, FCM) in order to evaluate the results and we comparatively analyze the obtained results on two case studies. This paper focuses on clustering with very large datasets and very many attributes of different types.